the long run, there is no distinction between arming ourselves and arming our enemies.” (p. 129) The Information Age is also the Dysinformation Age.
What can we do? We need to rethink our priorities with the help of the passionate but flawed analyses of Wiener, Weizenbaum, and the other serious critics of our technophilia. A key phrase, it seems to me, is Wiener’s almost offhand observation, above, that “these machines” are “helpless by themselves.” As I have been arguing recently, we’re making tools, not colleagues, and the great danger is not appreciating the difference, which we should strive to accentuate, marking and defending it with political and legal innovations.
Perhaps the best way to see what is being missed 1s to note that Alan Turing himself suffered an entirely understandable failure of imagination in his formulation of the famous Turing Test. As everyone knows, it is an adaptation of his “imitation game,” in which a man, hidden from view and communicating verbally with a judge, tries to convince the judge that he is in fact a woman, while a woman, also hidden and communicating with the judge, tries to convince the judge that she is the woman. Turing reasoned that this would be a demanding challenge for a man (or for a woman pretending to be a man), exploiting a wealth of knowledge about how the other sex thinks and acts, what they tend to favor or ignore. Surely (ding!)?, any man who could beat a woman at being perceived to be a woman would be an intelligent agent. What Turing did not foresee is the power of deep-learning AI to acquire this wealth of information in an exploitable form without having to understand it. Turing imagined an astute and imaginative (and hence conscious) agent who cunningly designed his responses based on his detailed “theory” of what women are likely to do and say. Top-down intelligent design, in short. He certainly didn’t think that a man, winning the imitation game, would somehow become a woman; he imagined that there would still be a man’s consciousness guiding the show. The hidden premise in Turing’s almost-argument was: Only a conscious, intelligent agent could devise and control a winning strategy in the imitation game. And so it was persuasive to Turing (and others, including me, still a stalwart defender of the Turing Test) to argue that a “computing machine” that could pass as human in a contest with a human might not be conscious in just the way a human being is, but would nevertheless have to be a conscious agent of some kind. I think this is still a defensible position—the only defensible position—but you have to understand how resourceful and ingenious a judge would have to be to expose the shallowness of the facade that a deep-learning AI (a tool, not a colleague) could present.
What Turing didn’t foresee is the uncanny ability of superfast computers to sift mindlessly through Big Data, of which the Internet provides an inexhaustible supply, finding probabilistic patterns in human activity that could be used to pop “authentic’- seeming responses into the output for almost any probe a judge would think to offer. Wiener also underestimates this possibility, seeing the tell-tale weakness of a machine in not being able to
take into account the vast range of probability that characterizes the human situation. (p.181)
° The surely alarm (the habit of having a bell ring in your head whenever you sce the word in an argument) is described and defended by me in Jntuition Pumps and Other Tools for Thinking (2013).
44
HOUSE_OVERSIGHT_016264